MySQL SELECT sum of a column, GROUP BY date failing when ONLY_FULL_GROUP_BY is enabled

Question

My Query:

SELECT      portfolio_id, 
            sum(shares_owned_value_usd) as sum_shares_owned_value_usd, 
            date_utc, 
            sum(dividends_received_value_usd) as total_dividends_received_usd,
            (sum(dividends_received_value_usd) over (order by date_utc range between interval 1 year preceding and current row)) as dividends_usd_ttm
FROM        stock_portfolio_historical_values
WHERE       portfolio_id = 5
GROUP BY    date_utc
ORDER BY    date_utc ASC

Error:

Error Code: 1055. Expression #3 of SELECT list is not in GROUP BY clause and contains nonaggregated column 'stock_portfolio_historical_values' which is not functionally dependent on columns in GROUP BY clause; this is incompatible with sql_mode=only_full_group_by

My table stock_portfolio_historical_values has a primary key named id , if this is needed to fix my query.

My Schema:

id	portfolio_id	ticker	shares_owned_value_usd	dividends_received_value_usd	date_utc
500	5	AAPL	300	15	2022-06-08
499	5	MSFT	1000	30	2022-06-08
498	5	AAPL	200	0	2022-06-07
497	5	MSFT	900	0	2022-06-07
.	.	.	.	.	.
403	5	AAPL	300	15	2022-03-05
402	5	MSFT	1000	30	2022-03-05
.	.	.	.	.	.
102	5	AAPL	300	10	2021-03-05
101	5	MSFT	1000	15	2021-03-05

Expected Results:

portfolio_id	sum_shares_owned_value_usd	total_dividends_received_usd	dividends_usd_ttm	date_utc
5	1300	45	90	2022-06-08
5	1100	0	45	2022-06-07
.	.	.	.	.
.	.	.	.	.

Note: This query works fine with ONLY_FULL_GROUP_BY disabled, but I am trying to get it to work with it enabled.

Answer 1

One option is separating the aggregation from the windowing operation, then join back the two results:

WITH cte AS (
       SELECT SUM(shares_owned_value_usd)        AS sum_shares_owned_value_usd,
              SUM(dividends_received_value_usd)  AS total_dividends_received_usd,
              date_utc 
       FROM stock_portfolio_historical_values
       WHERE portfolio_id = 5
       GROUP BY date_utc
)
SELECT SUM(dividends_received_value_usd) OVER(
              ORDER BY date_utc 
              RANGE BETWEEN INTERVAL 1 YEAR PRECEDING AND CURRENT ROW) AS dividends_usd_ttm,
       stock.date_utc
FROM      cte 
LEFT JOIN stock_portfolio_historical_values stock
      AND stock.date_utc     = cte.date_utc 
      AND stock.portfolio_id = 5

Does it work for you?